Automatic generation of nonlinear optimal control laws for broombalancing using evolutionary programming

摘要

The paper explores the use of mutatian operators with evolutionaryprogramming (EP) to automatically generate time optimal“bang-bang” type control laws for the three dimensionalbroom balancing (inverted pendulum) problem. EP produces a time optimalnonlinear control strategy that takes the state variables as input anddetermines the direction of the “bang-bang” force to beapplied. Preliminary results indicate that the control laws generatedare capable of generalizing over previously unseen input states andcompare well with nonlinear control laws that were generated using otherevolutionary computation methods